The main psychotic disorders, schizophrenia (SZ) and bipolar disorder (BD), continue to rank amongst the leading causes of disability worldwide largely because current clinical syndromal definitions are insufficient for treatment and prognosis because they are inadequately aligned with underlying pathophysiology. This proposal uses the Research Domains Criteria (RDoC) framework in order to define and validate biologically informed and clinically relevant neural phenotypes for psychotic disorders. Specifically, neuroimaging studies in patients with SZ and BD suggest that dysconnectivity within neural networks linked to the RDoC domains of perception, cognitive control and facial affect processing is central to the pathophysiology of psychosis. Further, abnormalities in these domains have been proposed to explain the clinical symptoms and cognitive deficits associated with psychotic disorders. Accordingly, our overall hypothesis is that abnormalities in effective connectivity, within domain-general neural networks of perception, cognitive control and facial affect identification, will detect biologically and clinically relevant neural phenotypes for psychosis. We present preliminary data that show that patients with SZ or BD can be classified into subgroups defined by their neural network architecture and that these neural phenotypes can be mapped onto clinical dimensions. Our results are based on estimates of effective connectivity from a dynamic causal model of the domain-general networks engaged in perception and cognitive control during working memory. The neural phenotypes we identified showed partial overlap between SZ and BD and were associated with symptom severity and clinical course. Based on this evidence, the aims of this proposal are (a) to expand our preliminary results in order to identify neural phenotypes for psychosis based on effective connectivity parameters derived from dynamic causal models of domain-general networks of perception, cognitive control and facial affect processing and test their reproducibility in two independent samples, (b) to define the association between the identified neural phenotypes and clinical dimensions of symptomatology and course, and (c) to determine their predictive value for treatment response. The proposal benefits from the use of dynamic causal modelling, which can infer causal interactions between brain regions underlying altered network dynamics, from testing the validity of our results based on their reproducibility and from assessing the therapeutic relevance of the identified neural phenotypes. Successful completion of the studies proposed in this application will improve our understanding of the clinical and prognostic significance of abnormal brain connectivity in psychosis, provide a scientific basis for therapeutic planning, and facilitate targeted etiological investigations and the development of new therapeutic approaches.